3. Experiments
The most important properties of an amplifier are its amplification as a function of input frequency (often called the frequency response), and the internally generated noise in the amplifier. To measure the frequency response one inserts a sinusoidal signal of known amplitude and frequency and then measure the output signals magnitude and phase compared to the input. The resulting curves are then often plot in a Bode diagram where the phase and gain are given as functions of frequency. The noise of an amplifier is often modeled as an equivalent voltage noise generator and current noise generator at the input of a noiseless amplifier [7] . To obtain the magnitude of the voltage noise generator, the input is grounded shorting the current noise generator. A noise spectrum is measured at the output and by dividing the spectrum by the frequency response of the amplifier the input voltage noise is obtained. By connecting a sufficiently large resistor between the signal input and ground terminals, the voltage noise contribution from the current noise generator flowing through the resistor, is much greater than the voltage generators contribution. However, a resistor has in itself a Johnson noise power proportional to the resistance. The equivalent noise current is now obtained by measuring the output voltage and dividing it by frequency response to get total noise voltage at the input. The current noise contribution can be separated from the Johnson noise of the resistor and the voltage noise generator since the noise components are uncorrelated. The result is divided by the resistance to get the input noise current. Only voltage noise was measured in this work.
3.1 Instrumentation
The frequency response was divided into two intervals, set by the measurement equipment. A Hewlett-Packard HP35665A Dynamic Signal Analyzer was used in the frequency range 10 Hz - 50 kHz, 50 kHz being the upper limit of the HP running in this mode. In the 50 kHz - 30 MHz range a Stanford DS345 Function Generator and a Fluke PM3328A Digital Oscilloscope controlled by a measurement computer were used. The Stanford is capable of producing an accurate sinusoidal signal up to 30 MHz while the Fluke can acquire signal frequencies up to 100 MHz and also calculate the RMS. voltage of a trace. The computer was a Macintosh running LabView software.
Noise spectra were taken by the HP which had built-in FFT analysis features. The upper frequency limit in this mode was 100 kHz. To get a good resolution the frequency range was divided into two intervals, 10 Hz - 1 kHz and 1 kHz to 100 kHz, each with 400 points. No noise spectra were taken above 100 kHz.
3.2 Methods
Experiments were carried out for two different MESFETs , with gate areas of 0.049 mm2 and 0.012 mm2. For each FET the frequency response and noise spectra were measured at different bias points. A bias point is defined by a double number, the MESFET drain current Id and its drain - source voltage Vds. The DC level at the cascode output, Vdc thus effects the bias point for the upper JFET, but this parameter was held constant at 7.5 V midway between 0 and 15 V.
At room temperature each bias point was set by first trimming the drain current to its desired value using the source trimmer. The DC voltage was then set by the drain trimmer to about 7.5 V to maximize the cascode dynamic range. Last, the drain-source voltage was set by the gate trimmer. It was found necessary to iterate the last two steps to close in on the desired bias point. The drain current was independent on the two voltages. The bias points were measured with a hand-held Fluke DMM. It has a 3 digit accuracy which is quite sufficient in this application. The drain current was deduced from the voltage drop across the 100 ohm resistance at the source of Q1 (discussed in the previous chapter ).
The amplifier was cooled by immersing the cold box in liquid helium at 4.2 K in a dewar. The dipstick could be locked at different heights above or below the liquid surface level to obtain different temperatures. Measurements were only conducted at 4.2 K however. When a MESFET is cooled, its output resistance drops. This means that its drain current is no longer independent on the drain-source voltages. Thus the drain current drifts when adjusting the drain-source voltage. As a consequence all three biasing steps had to be iterated to obtain the correct bias point instead of only the last two as in the room temperature case.
To check whether or not there were any free charges at 4.2 K, the bias was removed and the FET was allowed to cool for one hour. When the bias was returned, the FET conducted immediately indicating a substantial amount of free charges at 4.2 K in this sample.
To Top
To Table of Contents
Next Chapter